Optical characterization method

Abstract

This invention includes an optical characterization method capable of determining various characteristics of samples that preferably include naturally-occurring color deviations. In the preferred embodiment, this method is used in connection with food products and, particularly, potato products such as french fried potatoes. The method of this invention overcomes such disadvantages by generating from the RGB pixel values a sample-based color space that corresponds to the naturally-occurring color deviations of the samples. Preferably, the method includes generating HSI pixel values that correspond to the RGB pixel values and are defined with respect to an HSI color space having hue, saturation, and intensity coordinates. The sample-based representation of the color characteristics of the samples is determined on a pixel-by-pixel basis. As a result, color grades for the samples can be assigned according to the distribution of pixels within the sample-based representation. Moreover, defects or color variation, as well as the length distribution of the samples, may also be determined.

We claim: 
     
       1. An optical grading method for grading optical characteristics of samples with predominant colors of a substantially common hue, comprising: obtaining an RGB video signal representing the optical characteristics of the samples with red, green, and blue color component video signals;   generating from the RGB video signal an HSI representation with hue, saturation, and intensity image components;   identifying from the HSI representation the substantially common hue of the predominant colors of the samples;   generating from the HSI representation of the samples a sample-based representation in which the substantially common hue of the predominant colors of the samples forms a grade axis of the sample-based representation; and   assigning grades to the samples according to the position along the grade axis;   the RGB video signal including multiple pixels representing the optical characteristics of different portions of the samples and HSI and sample-based representations being generated for each of the pixels;   the multiple pixels of the RGB video signal corresponding predominantly to a surface in the HSI representation and generating the sample-based representation including applying the following equations to each pixel (H', S', I') in the HSI representation: ##EQU7## in which the sample-based representation is represented as a cylindrical coordinate system, x represents a normalized direction vector corresponding to a best fit line for the HSI pixel values on the surface in the HSI representation and including a point (H 0 , S 0 , 0) , and n h  is a normal vector to the surface in the HSI representation.